Author response: A neural network model of differentiation and integration of competing memories
Victoria J. H. Ritvo, Alex Nguyen, Nicholas B. Turk‐Browne, Kenneth A. Norman · 2024
What determines when neural representations of memories move together (integrate) or apart (differentiate)?Classic supervised learning models posit that, when two stimuli predict similar outcomes, their representations should integrate.However, these models have recently been challenged by studies showing that pairing two stimuli with a shared associate can sometimes cause differentiation, depending on the parameters of the study and the brain region being examined.Here, we provide a purely unsupervised neural network model that can explain these and other related findings.The model can exhibit integration or differentiation depending on the amount of activity allowed to spread to competitors -inactive memories are not modified, connections to moderately active competitors are weakened (leading to differentiation), and connections to highly active competitors are strengthened (leading to integration).The model also makes several novel predictions -most importantly, that when differentiation occurs as a result of this unsupervised learning mechanism, it will be rapid and asymmetric, and it will give rise to anticorrelated representations in the region of the brain that is the source of the differentiation.Overall, these modeling results provide a computational explanation for a diverse set of seemingly contradictory empirical findings in the memory literature, as well as new insights into the dynamics at play during learning. eLife assessmentThis paper presents important computational modeling work that provides a mechanistic account for how memory representations become integrated or differentiated (i.e., having distinct neural representations despite being similar in content).The authors provide convincing evidence that simple unsupervised learning in a neural network model, which critically weakens connections of units that are moderately activated by multiple memories, can account for three empirical findings of differentiation in the literature.The paper also provides insightful discussion on the factors contributing to differentiation as opposed to integration, and makes new predictions for future empirical work.